Of the 379 general practices across Wales, 23 general practices provided interest from 27 individuals. Of the 27 that expressed an interest, N = 19 individuals agreed to participate. Despite recruiting above our target of 15 participants, it was felt by the authors that thematic data saturation had not been achieved and four additional participants were interviewed, taking the total number of participants to N = 19. Of the eight that expressed an interest but did not participate, the reasons for not participating included: no response after follow-up email (n = 4), unable to coordinate time/date for interview (n = 3), and no financial incentives to participate (n = 1).
Thematic analysis
From the data, we identified six key themes. The most prominent theme identified was ‘coding challenges’, followed by ‘motivation to code’, and ‘making coding easier’. Other codes identified were ‘the daily task of coding’, ‘what and when to code’, and ‘coding through COVID’. There were frequent overlaps in the themes, for example, participants often talked about the problems with coding and making coding easier interchangeably. The theme of what and when to code was identified as two separate themes initially but on further analysis these were grouped together for ease of discussion and their close association in terms of content. Each theme is described with illustrative quotes.
Coding challenges
Finding the right code gave participants the most difficulties when using clinical coding, causing much frustration. It was the most commonly cited problem and consistently reported by clinical and non-clinical participants:
‘… for medicine reviews you have about twelve different ones [codes]. Well, we can’t use twelve different ones for searches or for claims for example.’
(P01, in-house pharmacy)
This difficulty often led them to look for generic codes, use free-text to elaborate on the chosen code, or even to give up searching:
‘It’s hard to find the right code sometimes, you end up not bothering to code it properly.’
(P27, GP)
‘If I can find the code within three clicks I will use it, if not, I will use the nearest available code that I think fits, which isn’t always the most accurate code.’
(P25, GP)
‘I saw [a] patient with a specific problem and I couldn’t find the code, neither could I find something related, so I ended up putting something generic like “complaining of a lump” as opposed to something I wanted to actually put in.’
(P16, GP)
For clinicians, time pressure negatively affected the quality of their coding. As one participant commented:
‘You’ve ten minutes to see a patient, seven to eight minutes of that is clinically. Which means you have a very limited amount of time to document and to keep your surgery on time. That is my biggest pressure when it comes to coding. I don’t have the time to search for the right code, there just simply isn’t.’
(P12, GP)
Patients presenting with multiple problems led to clinicians struggling with how to capture the different problems in meaningful code:
‘It’s much easier to put everything down under one code, so if the patient comes in with a list of issues, say three or four problems … it’s easier for me to list them, one, two, three, four, under the same code. Rather than finding the correct code [for each individual problem].’
(P15, GP)
Miscoding was a cause of concern for all participants and the implications this had on patient care. One of the most cited examples was around coding incorrectly, or the codes not being added to the record for a significant diagnosis, such as cardiovascular disease. Participants discussed the potential consequences of this, such as not being on a disease register, not having regular reviews of the condition, and not being on appropriate or correct medication. Concerns were also expressed by participants about the potential knock-on effect of miscoding in a patient’s medical record when it came to administrative work, such as writing insurance or medical reports.
With increased access to medical records by patients, participants also expressed concerns around inconsistencies in coding of information:
‘If the patient wants their own medical history they might say well I’ve got x, y, and z and it’s not there [in their medical record].’
(P02, administrative)
‘I can think of one case in particular where a patient ended up with a diagnosis of schizophrenia, because essentially a patient’s mum had rung out of hours and said “my son’s behaving unusually, like my aunty who had schizophrenia” and the admin clerk just saw the word schizophrenia and coded the patient as schizophrenic. It wasn’t until I was doing a DVLA [Driver and Vehicle Licensing Agency] medical on him and I thought, well he’s not on any medication, this is a bit unusual and so I looked at the notes, and this is where the code has come from. So it’s something that if done incorrectly can create problems.’
(P25, GP)
Motivation to code
Overall, both clinical and non-clinical participants felt there was value to using clinical coding in their day-to-day work. Good clinical coding was felt to facilitate good care, improve patient safety, facilitate optimal workflow, allow consistent record keeping, and help with quality improvement and audit. Participants generally acknowledged the importance of clinical coding in receiving financial reimbursement for their work, for example, with enhanced services, QAIF, and previously QOF. However, there was a clear difference in the views of GPs and other employment roles. GP partners felt the financial requirements of coding were important while locums and salaried GPs felt the financial element was of little importance to them, highlighting the added responsibilities of being a partner and ensuring financial stability of a practice. Administrative staff showed understanding of the importance of accurate and consistent clinical coding and its impact on finances within the practice. This was summed up well by an administrative participant who had moved from a health board-managed practice to an independent contractor practice:
‘I see a bigger picture to it. There’s more of a reason for it now than I ever thought before … before I just wanted to do a good job and I was a bit smug that I knew how to do these codes and no one else did, whereas now I’m like, you should do these because of this reason [financial].’
(P07, administrative)
The use of free-text versus coding generated interesting differences between clinical and non-clinical staff. This spoke of the different motivations between members of the general practice team when it comes to clinical coding. When discussing their use of free-text, one clinician commented:
‘I find personally that kind of just freeform writing is where I’m going to convey the information to the next clinician. As a result, I think I’m probably pretty lax in terms of my choice of codes.’
(P12, GP)
Despite demonstrating motivation to want to code as accurately and consistently as possible, clinical participants did not feel that they were able to transfer the information patients were telling or giving them into codes. This also fitted with the time pressures that clinical participants mentioned in the ‘coding challenges’.
This clinician view contrasted with some of the comments made by administrative staff, who wanted information that was written as free-text to be coded as much as possible, demonstrating high levels of motivation to complete the coding as accurately as possible:
‘You can free-text, but that won’t come up in a report, or if a GP is looking for it, and it was twelve months ago.’
(P07, administrative)
‘We get summaries like the code is “had a chat to patient”, and then all free-typed is “patient has been diagnosed with prostate cancer stage four” and it is like none of this is on their medical record! But if you look at their record they’ve had thirty encounters of “had a chat with the patient”.’
(P04, administrative)
Making coding easier
Some participants had preference for particular EHR systems based on ease of use. Others reported coding was easier if they had good administrative support and local experts within a practice, being able to adapt the EHR system to make certain codes easier to find and use. Templates were sometimes seen as helpful for clinical staff. Templates are tools embedded into the EHR that allow consistent use of codes for a particular clinical problem or situation, without the need to search for these codes. Clinicians commented that templates enabled key codes to be quickly accessed, though were not always intuitive.
Administrative staff found templates helpful when preparing claims for services as it was easier to run reports, but they often found that they had to cajole and encourage clinical staff to use them:
‘It’s a hard sell when you want them [clinicians] to take it on [templates]. I’ve just built a minor op one [template for minor operations/surgery] … I‘ve done it completely selfishly as the codes weren’t being picked up when doing the claims. So, I was like, let’s fix the problem, let’s create a template. The doctor liked writing the way she did, but now she’s realised there’s a tick list she’s all for it because it saves her time.’
(P04, administrative)
The daily task of coding – ‘all day, every day’ (Participant [P] 18, administrative)
Regardless of their role, participants reported using clinical coding on a daily basis, typically multiple times in a day. The daily task of coding was a theme that permeated through all the other themes, and one that bounds the others together. We have not delved deeply into this theme here but reference it through the other themes. The quote above gives a feel for how participants viewed the monotony of clinical coding.
What and when to code?
Participants discussed dilemmas in relation to what and when to code. Clinicians described how they preferred to code a patient’s presenting complaint or symptom and/or signs, and felt uncomfortable in coding a definitive diagnosis:
‘I feel it comes down to symptoms versus diagnosis … with undifferentiated problems I almost certainly code symptoms. Even if someone comes in with a known diagnosis of something, but their symptoms possibly relate to it or possibly doesn’t, I still find I code the symptoms because I feel it is quite bold to code a definitive diagnosis for something without being sure.’
(P16, GP)
There were, however, examples when clinicians would code a diagnosis, for example, tonsillitis or otitis media. These were described as ‘straightforward consultations’, in otherwise healthy individuals.
Clinical coding was also used in novel ways by participants, for example, to flag patients who might need a different approach to the usual 10-minute consultation:
‘There’s a code called “multiple problems”. I will often use that if I know it’s a patient who would regularly come in with a list of problems. I always code that as a separate one, so that if I am searching it, and it says multiple problems, I know that that’s a patient who will either need a double appointment or will need to be closed down quite early on to keeping things specific.’
(P25, GP)
‘So if I put in the code “had a chat to patient”, that means there’s something else perhaps that’s not medically related — it’s just a prompt to me to ask.’
(P25, GP)
These examples were the exception rather than the norm, but presented interesting approaches to using clinical coding as aide-mémoires and beyond the original purposes of clinical coding.
Some clinicians described that they coded after the consultation if the consultation was face-to-face. This allowed them to give the patient their full attention. Choosing a code and typing while the patient was present was considered a distraction. However, clinical signs such as blood pressure and weight, where it was described as a simple ‘click of a button’, were often coded during consultations. Low-complexity consultations and telephone triage were often coded during the consultation as it was felt this had less impact on communication with the patient.
On an administrative level there were examples of innovative ways of working, such as surgeries grouping together to undertake the administrative task of coding, creating ‘specialist clinical coders’, whose sole role was to code clinical information. They were reported to provide a consistent approach to coding and summarising information across the practices. This required effort and commitment from GP partners, who supervised this initially, to ensure the quality of the coding, but with time this reduced the burden of administrative coding for GPs.
Coding through COVID
Though COVID-19 represented a small proportion of the discussions with participants, it was evident that people changed their coding behaviour owing to the pandemic. They reported that use of generic codes such as ‘telephone encounter’ became more prominent during this period. Participants felt this had a negative effect on data quality, with large sections of patient contact being coded under generic administrative codes, rather than meaningful clinical information relating to patients’ health:
‘During COVID I worked in more than one place where everything went to telephone consults, [therefore] everyone started coding [consultations] as “tele-consult” and lots of people are still doing that. So, no matter what the consultation is about, the code will say “telephone call” or “home visit”, which doesn’t give you any clue as to what’s happening [with the patient].’
(P16, GP)
There were also external pressures to quantify and measure the different methods practices were using for patient contact:
‘The type of coding that they [clinicians] were using probably did change because there were far more telephone consultation [and] video consultation, [therefore] those types [of] codes [“telephone call to patient” and “video call”] were being introduced more [because] we were being asked to prove the different types of contact we were having [with patients].’
(P08, administrative).